Penny Lee
Penny Lee Co-founder of LRAI, the first AI search engine for legal research in Australia.

How We Created the First AI-Driven Legal Research Service with Guaranteed No Hallucinations

How We Created the First AI-Driven Legal Research Service with Guaranteed No Hallucinations

Accurate legal research is crucial for building strong cases and providing sound legal advice. However, traditional research methods are time-consuming, and existing AI tools often fall short due to issues like hallucinations, where the AI generates misleading or incorrect information. This not only undermines trust in AI but also poses significant risks in the legal field.

Recognizing this challenge, we set out to develop the first AI legal research service that eliminates hallucinations structurally, so that we get the best of both worlds – the time-saving advantages of AI technology and the certainty of information provided by traditional legal research methodologies.

Large language models, which many AI tools rely on, are known to hallucinate, especially without grounding information (i.e., providing the context to the large language model). Even when the context is provided (for e.g. Retrieval Augmented Generation or RAG), large language models can still hallucinate. Furthermore, extracting the correct context from millions of cases and legislation at a reasonable cost is a significant challenge. This is why almost all other legal AI developments fall short - their aim is always to produce a chatbot!

Knowing this, we intentionally did not produce a chatbot as our main feature. Instead, we focused on using AI techniques to do an extensive upgrade of the traditional keyword-based search function. After all, the main goal of legal research is to find specific case law and legislative clauses to reference in constructing a strong argument, not to chat with a chatbot.

We structured our service such that AI is used for contextual understanding and analysis, but similar to traditional keyword searches, all our search results come from actual databases of case law and legislation and not from a large language model. Hence, hallucinations are impossible in this type of structure. Of course, due to the new AI techniques we incorporated, our search results are significantly more relevant to your queries compared to a simple keyword search which has no understanding of context or meaning.

Developing this tool has been a rewarding journey, driven by our commitment to improving legal research. We believe that with the right tools, legal professionals can achieve greater accuracy and efficiency, ultimately benefiting their clients and the broader legal community.

We invite you to experience the future of legal research at lrai.com.au.

comments powered by Disqus